New
bitsandbytes-foundation/bitsandbytes 0.50.2 — 4·8-bit quantization
bitsandbytes is an MIT-licensed library that quantizes PyTorch model weights to 4·8 bits to reduce the memory needed for inference and training; the latest release is 0.50.2.
bitsandbytes-foundation/bitsandbytes
bitsandbytes is an MIT-licensed library that quantizes PyTorch model weights to 4·8 bits to reduce the memory needed for inference and training; the latest release is 0.50.2.
View on GitHub ↗- stars
- ★ 8.5K
- Forks
- 939
- Language
- Python
- License
- MIT
- Latest release
- 0.50.2
Posts
1No matching stories.